This is the fifth revision of my xfstest adding a test for the new mmap() MAP_SYNC flag. The previous version can be found here: https://lists.01.org/pipermail/linux-nvdimm/2017-November/013329.html Changes since v4: - Test for dm-log-writes + DAX support in the spirit of _require_scratch_dax() instead of relying on specific module version number. (Eryu) - Print out the missing mark name in the error output of _log_writes_replay_log(). (Eryu) - Go back to using the output from "du -sh" to test our file block allocations. The "stat -c %s" I was using in v4 was broken if the file existed and only had partial block allocations, and the number of blocks reported by "stat -c %b" can vary based on the block size. - Move the test number to 999 to avoid rebasing collisions. Feel free to renumber as necessary when merging. I've updated the links at the bottom of this mail to point to the most recent version of my xfsprogs changes and to the current version of this series. --- The purpose of this series is to exercise the new MAP_SYNC mmap() functionality [1]. It adds a test which uses dm-log-writes to try and replay filesystem metadata operations for a file that is being written via mmap(). If MAP_SYNC is active the dm-log-writes replay will recreate the file's block allocations and you'll end up with a test file which is a known size. If MAP_SYNC is not active the metadata writes will most likely be lost and the replay will either fail to create the test file at all or it will may be smaller. In all of my testing the file simply doesn't exist on the replay if MAP_SYNC is ommited. This test relies on a kernel with both the MAP_SYNC mmap() functionality and the DAX enabling in dm-log-writes. These both appear in kernel v4.15-rc*. For ease of testing I've posted a kernel that is v4.14 plus just those two patch series [2]. This test also relies on xfsprogs having support for MAP_SYNC and for dm-log-writes. I've posed a tree adding that support [3]. This xfsprogs series is still under review so if the xfs_io interfaces change this test will of course need to be updated as well. Lastly, I've also posted a working version of this series [4]. [1]: https://lists.01.org/pipermail/linux-nvdimm/2017-November/013164.html [2]: https://git.kernel.org/pub/scm/linux/kernel/git/zwisler/linux.git/log/?h=map_sync_dm_log_writes [3]: https://git.kernel.org/pub/scm/linux/kernel/git/zwisler/xfsprogs-dev.git/log/?h=map_sync_v2 [4]: https://git.kernel.org/pub/scm/linux/kernel/git/zwisler/xfstests-dev.git/log/?h=map_sync_test_v5 Ross Zwisler (2): dm-log-writes: only replay log to marks that exist generic: add test for DAX MAP_SYNC support common/dmlogwrites | 24 +++++++++++++++ tests/generic/999 | 82 +++++++++++++++++++++++++++++++++++++++++++++++++++ tests/generic/999.out | 3 ++ tests/generic/group | 1 + 4 files changed, 110 insertions(+) create mode 100755 tests/generic/999 create mode 100644 tests/generic/999.out -- 2.14.3 -- To unsubscribe from this list: send the line "unsubscribe linux-xfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
